GAN Based Handwritten Digit Classification Using CNN^*

By utilising two competing networks, Generative Adversarial Networks have emerged as a ground-breaking method for producing synthetic data. In order to demonstrate how GANs may enhance training datasets, enhance model generalisation, this study investigates the application of GANs in handwritten digit production and categorisation. Mode breakdown and training instability are still issues despite their potential. The study employs Frechet Inception Distance (FID) to assess the quality of generated images and shows that GAN-derived data improves classification accuracy when combined with Convolutional Neural Networks (CNNs). To overcome current constraints and expand GANs’ capabilities to more complicated and noisy datasets for wider applications, more research is necessary.